It　is　possible　for　IT　service　providers　to　provide　computing　resources　in　an　pay-per-use　way　in　Cloud　Computing　environments.　At　the　same　time,　terminal　users　can　also　get　satisfying　services　conveniently.　But　if　we　take　only　execution　time　into　consideration　when　scheduling　the　cloud　resources,　it　may　occur　serious　load　imbalance　problem　between　Virtual　Machines　(VMs)　in　Cloud　Computing　environments.　In　addition　to　solve　this　problem,　a　new　task　scheduling　model　is　proposed　in　this　paper.　In　the　model,　we　optimize　the　task　execution　time　in　view　of　both　the　task　running　time　and　the　system　resource　utilization.　Based　on　the　model,　a　Particle　Swarm　Optimization　(PSO)　-　based　algorithm　is　proposed.　In　our　algorithm,　we　improved　the　standard　PSO,　and　introduce　a　simple　mutation　mechanism　and　a　self-adapting　inertia　weight　method　by　classifying　the　fitness　values.　In　the　end　of　this　paper,　the　global　search　performance　and　convergence　rate　of　our　adaptive　algorithm　are　validated　by　the　results　of　the　comparative　experiments.　©　2012　Springer-Verlag.